Exploring the relationship between average speed, speed variation, and accident rates using spatial statistical models and gis

TLDR: In this paper, the authors investigated the relationship between average speeds, speed variations, and accident rates using two advanced statistical models: (1) a nonspatial random-effects negative binomial model and (2) a spatial Poisson-lognormal model using a full hierarchical Bayesian model.

Abstract: The primary objective of this article is to contribute to the debate on the relationship between average speeds, speed variations, and accident rates. This is achieved by the use of two advanced statistical models: (1) a nonspatial random-effects negative binomial model and (2) a spatial Poisson-lognormal model using a full hierarchical Bayesian model to explore the relationship. Disaggregated segment-based traffic, road geometry, and accident data from 266 road segments including 13 different motorways (including the M25 motorway) and 17 different trunk A-class roads around London from 2003 to 2007 are used in the analysis. GIS tools are used to achieve the appropriate data and to derive the weight matrix among neighboring segments that is necessary for the spatial model. The results suggest that average speeds are not associated with accident rates when controlling for other factors affecting accidents such as traffic volume, road geometry (e.g., grade and curvature), and number of lanes. However, speed variation is found to be statistically and positively associated with accident rates. A 1% increase in speed variation is associated with a 0.3% increase in accident rates, ceteris paribus. The results for all other factors are found to be consistent with existing studies. Policy implications of the findings are then discussed.